Improve performance of `TimelineFieldsResource` endpoint

[jkppr]

In PR #3182 we introduced a new API endpoint TimelineFieldsResource to gather unique field names per timeline to improve the UX of the visualizations dialog.

https://github.com/google/timesketch/pull/3182/files#diff-bb4334f2502d8d96c5689e099dd1440470e425b43a01d8221ebfaabd18511862R538

Currently, the endpoint retrieves the unique fields by:

1. Aggregating all data types within the timeline.
2. For each data type, querying for a single event to get the fields.

This approach can be inefficient, especially for timelines with many data types.

This issue is to track research into a more efficient way of gathering those fields. Possible approaches to investigate include:

• Using OpenSearch field capabilities: Explore if OpenSearch provides a more direct way to retrieve unique fields for a given index and data type.
• Caching field information: Investigate caching the field information to avoid redundant queries.
• Optimizing the aggregation: Analyze if the aggregation step can be optimized or if it can be avoided altogether.

Tasks:

• [ ] Research alternative approaches for retrieving unique fields.
• [ ] Benchmark the performance of different approaches.
• [ ] Implement the most performant solution.

Related PRs:

• #3182